概率积分法模型参数全局灵敏度分析

Global Sensitivity Analysis for Probability Integral Method

采用多元逐步回归法对矿区变形预测模型——概率积分法模型各参数进行逐步回归分析,由回归顺序确定模型各参数的全局灵敏度。以淮南矿区1212(3)工作面为例,采用@RISK模拟计算,结果表明,下沉系数对模型起最关键的作用,灵敏度最高,与模型的相关系数为0.85,其次为主要影响角正切和拐点偏移距,相关系数为0.09和0.03,开采影响传播角的灵敏度最低,相关系数为0.004。该研究结果为模型参数的选择提供了数学层面的依据,在应用概率积分法模型进行开采沉陷预计时,应准确确定灵敏性高的下沉系数和主要影响角正切值这两个参数。

Multiple stepwise regression analysis approach was of probability integral model in this study. With work facing 1212 simulated by @ RISK have showed that the subsidence coefficient 0.85. Followed by deviation of inflection point and the tangent of 0.03 and 0.09 respectively. The influence propagation angle plays used to assess the global sensitivity of the parameters （3） in Huainan mining areas as example, the results plays a dominant role in the model, the coefficient is main influence angle, their correlation coefficients are a minimum role in this model, its coefficient is 0.004. The research resuhs provide a mathematical basis for the choice of parameters. When using probability integral method to predict the subsidence, the subsidence coefficient and the tangent of main influence angle should be given as accurate as possible.